1st Workshop on Formal Verification of Machine Learning (WFVML 2022)
Co-located with ICML 2022, at Baltimore Convention Center
Baltimore, Maryland, United States (Physical Workshop)
About This Workshop
When machine learning-based building blocks become widely available for complex and critical systems such as autonomous vehicles, medical devices, or cyber-security systems, their behavior must also be exactly characterized to ensure the high assurance of the entire system. Most existing research treats a machine learning model such as a deep neural network as a black box and uses simple empirical metrics such as accuracy to quantify their performance. However, accuracy alone is not sufficient to assure that the model always obeys even basic specifications. Formal verification algorithms for machine learning aim to formally prove or disprove desired specifications of a machine learning model. Some common specifications include safety, fault tolerance, fairness, robustness and correctness.
The aims of this workshop are:
Bring together researchers interested in the emerging field of machine learning verification from a broad range of disciplines (such as computer-aided verification, programming languages, robotics and control, computer security, and optimization) with different perspectives to this problem;
Raise awareness of the importance of formal verification methods in the machine learning community and stimulate more research that can tackle open challenges on the verification of real-world applications such as robotics and control;
Chart out important and promising future directions towards novel verification algorithms with better scalability and applicability.
Our workshop features 6 invited speakers spanning a quite diverse research background, including robotics, programming languages, optimization and computer security.